Dynamic and static stress synchronous measurement method based on strain gauge

Abstract

The application discloses a dynamic and static stress synchronous measurement method based on a strain gauge, and relates to the technical field of stress measurement. The method comprises the following steps: an edge gateway performs the following steps: intercepting an aliasing strain sequence with an absolute time stamp according to an overlapping time window, extracting and temporarily storing an extremely low frequency static load baseline, and forming a pseudo alternating sequence; generating a forward continuation sequence according to the tail end envelope trend of the pseudo alternating sequence and in combination with the main frequency of a measured entity structure; implementing bidirectional time reversal filtering on the total sequence and eliminating the forward continuation sequence to obtain a zero phase shift alternating sequence; resetting the static load baseline and inputting a temperature sequence into a lag constraint state space model to obtain a static creep reference; and superimposing the zero phase shift alternating sequence and the static creep reference according to the absolute time stamp to output a synchronous stress sequence. The method can reduce time window boundary distortion, inhibit group delay dislocation and temperature drift, and improve dynamic and static stress reconstruction consistency and fatigue evaluation reliability.

Description

Technical Field

[0001] This invention relates to the field of stress measurement technology, specifically to a method for simultaneous measurement of dynamic and static stress based on strain gauges. Background Technology

[0002] Strain gauges, attached to the surface of a component, detect changes in resistance to represent the deformation state of the component under stress. In engineering, they commonly employ Wheatstone bridges, signal conditioning, voltage acquisition, and back-end processing. They are frequently used in long-term service structures such as bridges, tunnels, wind turbine towers, railway tracks, and pipelines, as well as in tensile sensors, material testing, and stress analysis of industrial equipment. Because the monitored objects often simultaneously experience slowly changing static forces, superimposed dynamic vibrations and temperatures, installation conditions, and drift due to long-term service, mainstream technologies typically require simultaneous acquisition and processing of static, dynamic, and temperature-related errors.

[0003] US Patent document US11740268B2 discloses a predictive zero-phase filtering method. This document addresses the frequency component separation of real-time signals by acquiring historical measurements of the signal to be processed and obtaining predicted values. Both historical measurements and predicted values ​​are simultaneously input into a non-causal zero-phase filtering algorithm. The processing chain is as follows: an input time window is created near the current time; the input sequence undergoes a first filter; the first filter result is time-reversed for the first time, followed by a second filter; the second filter result is time-reversed for the second time, and the waveform value corresponding to the current time is output. The purpose of this document is to compensate for the inability to directly use future values ​​in real-time filtering by predicting future values, thus enabling zero-phase filtering to be applied online for frequency separation of a single input signal.

[0004] However, existing methods still have problems when using strain gauges for long-term online monitoring of stressed components or for continuous measurement by tensile sensors. First, the strain gauge output is not a simple frequency signal; it contains a large static base value, slowly changing low-frequency components, rapidly changing vibration components, and thermal drift caused by temperature changes. When the mixed signal is passed through a prediction and zero-phase filter chain in the form of a time window, the low-frequency base value and slowly changing drift directly participate in the prediction and filtering near the boundary, and the window has a greater influence on the waveform. Second, temperature changes do not occur within a certain time. Heat transfer, material thermal inertia, and slow-changing creep of the structure create multiple sources of change in the low-frequency range. It is difficult to distinguish which changes come from the actual force and which come from temperature hysteresis or long-term drift when using a single input signal for prediction and zero-phase filtering. Third, in applications that require both the preservation of dynamic vibration phase relationships and the acquisition of static average forces, if the mixed signal is uniformly filtered, the resulting frequency components can be used for general frequency selection, but for synchronous measurement of dynamic and static stress, problems such as low-frequency base value entrainment, residual thermal drift, and unstable time correspondence still exist.

[0005] Therefore, in the online measurement of mixed strain gauge signals, it is necessary to solve how to separate and reconstruct mixed strain signals with static stress, dynamic vibration and temperature-related drift while maintaining the time correspondence. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention provides a strain gauge-based method for synchronous measurement of dynamic and static stress. Based on the tail envelope tendency of the pseudo-alternating sequence and combined with the dominant frequency of the measured entity structure, a look-ahead extension sequence is generated. The total sequence undergoes bidirectional time-inverse filtering, and the look-ahead extension sequence is pruned to obtain a zero-phase-shift alternating sequence. The static load back-baseline is reset and combined with the temperature sequence, input into the hysteresis constraint state-space model to obtain a static creep reference. The zero-phase-shift alternating sequence and the static creep reference are superimposed on the same track according to absolute timestamps to output a synchronous stress sequence. This method can reduce time window boundary distortion, suppress group delay misalignment and temperature drift, and solve the technical problems described in the background art.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] The method for synchronous measurement of dynamic and static stress based on strain gauges, executed by an edge gateway, includes: acquiring aliased strain sequences with absolute timestamps and truncating them according to overlapping time windows; extracting and temporarily storing the extremely low frequency static load negative baseline for each time window; and subtracting the static load negative baseline from each time window to form a pseudo-alternating sequence.

[0011] Based on the tail envelope tendency of the pseudo-alternating sequence and combined with the main frequency of the tested entity structure, a look-ahead extension sequence is generated. The total sequence containing the look-ahead extension sequence is subjected to bidirectional time-inverse filtering and the look-ahead extension sequence is removed to obtain a zero-phase-shift alternating sequence.

[0012] The static load back baseline is reset, and the temperature sequence is input into the hysteresis constraint state space model to separate thermal drift and mechanical creep, thus obtaining the static creep baseline.

[0013] The zero-phase-shift alternating sequence is superimposed on the static creep reference along the same track according to the absolute timestamp, and the synchronous stress sequence is output.

[0014] Furthermore, the edge gateway extracts the static load baseline in the central area after removing the first and last boundary samples in each time window;

[0015] The static load negative baseline of the previous time window is transitioned to the static load negative baseline of the current time window according to the sample point number for the overlapping area shared by adjacent time windows, forming a transition baseline; the mixed strain sequence is stripped point by point according to the transition baseline to form a pseudo-alternating sequence.

[0016] Furthermore, the edge gateway extracts the training segment from the tail of the pseudo-alternating sequence and generates a look-ahead extended sequence based on the tail envelope trend; it concatenates the pseudo-alternating sequence of the overlapping area of ​​the previous time window, the pseudo-alternating sequence of the current time window, and the look-ahead extended sequence into a total sequence; it performs forward filtering and inverse filtering after time reversal on the total sequence using the same set of digital filtering coefficients; and it retains only the samples within the boundary of the current time window as the zero-phase-shift alternating sequence.

[0017] Furthermore, the edge gateway backfills the static load backing baseline into a reset static load sequence according to the absolute timestamp, and aligns the temperature sequence with the reset static load sequence to the same grid; the reset static load sequence and temperature sequence are input into a hysteresis constraint state space model containing mechanical creep state, thermal wandering residual state and thermal inertia memory state; and the static creep baseline is extracted from the output state of the hysteresis constraint state space model.

[0018] Furthermore, the edge gateway superimposes the zero-phase-shift alternating sequence and the static creep reference on the same track according to the absolute timestamp, including: writing the time window number, time window start timestamp, time window length and sampling interval into the page header of the zero-phase-shift alternating sequence and the page header of the static creep reference, respectively; reconstructing the time index of the two sample points based on the page header fields; and performing the same track superposition only on sample point pairs whose time difference is not greater than the synchronization tolerance.

[0019] Furthermore, when the edge gateway outputs the synchronous stress sequence, it encapsulates the synchronous stress sequence into a data frame. The frame header of the data frame includes the measurement point identifier, time window number, time window start timestamp, sampling interval, and alarm status word. The frame body includes the sample point array of the synchronous stress sequence and the corresponding peak and valley indexes, and is output continuously in page order.

[0020] Furthermore, the edge gateway establishes a mismatch record area for sample pairs that fail the synchronization tolerance judgment; the mismatch record area records the time window number, sample number, zero phase shift alternating sequence sample time index, static creep reference sample time index, and sample time difference; the edge gateway skips the sample pairs corresponding to the mismatch record area within the current time window and does not write them into the synchronization stress sequence.

[0021] Furthermore, the length of the overlapping time window is 0.5 to 0.9 times the first-order low-frequency natural period of the tested entity, and the overlap ratio of the overlapping time window is 0.25 to 0.5. After the edge gateway excludes the first and last boundary samples in the central area of ​​each time window according to the predetermined boundary elimination ratio, it then extracts the extremely low frequency static load back baseline.

[0022] Furthermore, the edge gateway determines the length of the look-ahead extension sequence according to the first-order low-frequency natural period of the tested entity and the sampling interval. The time span corresponding to the look-ahead extension sequence is 1.25 to 1.5 times the first-order low-frequency natural period. The edge gateway extracts a training segment with a length of 3 to 6 times the regression order from the end of the pseudo-alternating sequence to generate the look-ahead extension sequence.

[0023] Furthermore, the edge gateway includes a real-time clock module, a double-buffered static random access memory, and a direct memory access scheduling unit; the direct memory access scheduling unit writes adjacent time windows to the two memory areas during half-transmission interruption and write-full interruption, respectively; the edge gateway alternately performs static load back baseline extraction, look-ahead extension sequence generation, and co-track overlay between the two memory areas.

[0024] (III) Beneficial Effects

[0025] This invention provides a method for simultaneous measurement of dynamic and static stress based on strain gauges, which has the following advantages:

[0026] By extracting and temporarily storing the aliased strain sequence through the edge gateway, the static load negative baseline is first obtained and then a pseudo-alternating sequence is formed. The subsequent dynamic chain and the low-frequency static load channel are separated first to avoid the static load negative baseline directly entering the bidirectional time-inverse filter and causing cliff distortion at the time window boundary, thus ensuring that the starting point of dynamic and static component reconstruction is consistent.

[0027] Based on the tail envelope tendency of the pseudo-alternating sequence and combined with the dominant frequency of the tested entity structure, a look-ahead extension sequence is generated. The total sequence is then subjected to bidirectional time-inverse filtering, and the look-ahead extension sequence is pruned. The filtering boundaries at the beginning and end of the time window are constrained within a controllable range, resulting in a zero-phase-shift alternating sequence corresponding to the original time scale. The static load negative baseline is reset and simultaneously input into the hysteresis-constrained state-space model along with the temperature sequence. This separates thermal drift, thermal inertia memory, and mechanical creep within the same low-frequency channel, avoiding misjudging temperature hysteresis as static stress and outputting a static creep baseline.

[0028] By unifying the constraints of zero-phase-shift alternating sequences and static creep benchmarks using absolute timestamps, and performing same-track superposition after same-track determination, dynamic and static components no longer depend on arrival order splicing, eliminating false peaks and valleys and false load histories caused by cross-window mismatch and asynchronous superposition. Static load baseline extraction, look-ahead extension sequence generation, hysteresis constraint state-space model decoupling, and same-track superposition are performed continuously using a single edge gateway. The data link is closed and the object is unique, allowing direct output of synchronous stress sequences, eliminating time-series drift and interface ambiguity caused by multi-entity division of labor. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the field deployment structure of the strain gauge-based synchronous dynamic and static stress measurement system according to an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of the four-step closed-loop overall process of the strain gauge-based synchronous measurement method for dynamic and static stress according to an embodiment of the present invention.

[0031] Figure 3 This is a schematic diagram of the layering mechanism of overlapping windowing, central extraction, and static load track separation in an embodiment of the present invention;

[0032] Figure 4 This is a schematic diagram of the phased processing of look-ahead extension and bidirectional time-reverse washing in an embodiment of the present invention;

[0033] Figure 5 This is a functional block diagram of static load reset and thermal hysteresis decoupling in an embodiment of the present invention;

[0034] Figure 6 This is a block diagram of the integrated output chain for same-track stitching, stress conversion, and rainflow closed-loop output in an embodiment of the present invention;

[0035] Figure 7 This is a schematic diagram showing the relationship between synchronous stress output and fatigue closed-loop in an embodiment of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] Please see Figures 1-7 This invention provides a method for simultaneous measurement of dynamic and static stress based on strain gauges, comprising:

[0038] This implementation first uses the aliased strain sequence with absolute time scale. Cut into overlapping data windows, and then set the static load baseline corresponding to the macroscopic constant load. The data is stripped from and stored inside the window, allowing subsequent forward extension, bidirectional time-reverse scrubbing, and thermal creep reset to share the same data base.

[0039] Step 1: While preserving the original absolute timescale, process the aliased strain sequence. Drag the entire window to find the average static load baseline First, the signal is strongly isolated, and the remaining signal is transformed into a pseudo-alternating sequence for subsequent inverse wave processing. The following steps are all performed by the edge gateway.

[0040] Strain acquisition bridge outputs aliased strain sequence The clock module outputs an absolute time stamp, and the temporary register area stores the static load baseline. At measurement points on bridge cables, wind turbine towers, or rolling mill columns, long-term constant loads will raise the overall sampled values ​​and cause them to drift slowly. If subsequent forward and inverse filtering are directly applied to such sequences, the truncation boundary will misinterpret the macroscopic constant load as a sudden change, and the boundary distortion will spread in the reverse direction along the inverse wave, causing the group delay compensation to lose its basis.

[0041] Therefore, step one is not a simple mean correction, but rather splits the same observation into two paths with different physical meanings: static load baseline. As a low-frequency carrier for subsequent reset, the pseudo-alternating sequence As an alternating carrier for subsequent cleaning. Only by first removing the former from the aliased strain sequence. Only by completely removing the base and storing it according to the original time scale can subsequent forward extensions not be pulled by the baseline step, and only then can the thermal hysteresis stripping in step three have objects to be backfilled.

[0042] First, advance the overlapping data windows according to the absolute time scale to retain common samples between adjacent windows; then extract the static load baseline only in the central area of ​​the window. Actively avoid the first and last boundary sampling points; then, within the overlapping section, let the previous window static load bear the baseline. Smooth transition to the current window static load baseline To avoid creating additional steps in the baseline removal process; finally, statically load the baseline. Push the pseudo-alternating sequence into the temporary storage area. It is then sent to the next step along with the original time stamp.

[0043] In a preferred embodiment, the tested entity is a bridge suspension cable. The sensing front end uses a constantan foil grid resistance strain gauge, which is fixed to a steel pad near the anchor head by a two-component epoxy adhesive layer. The lead wire is a shielded twisted pair cable connected to the analog-to-digital converter board. The analog-to-digital converter board and the computing board are placed in the same protective enclosure. The actions that can be directly observed on site are: the sample points are first written into a dual-zone annular buffer, and then the computing board cuts out overlapping windows; the slowly rising part of the central area is removed and stored separately, and the remaining sequence is converted into a waveform skeleton oscillating around the zero line, and there are no longer abrupt steps at the beginning and end of the window.

[0044] The edge gateway continuously advances overlapping data windows based on absolute timescales, rather than temporarily slicing them according to the arrival order of software threads. The reason is that subsequent reverse traversal requires a handover zone shared by the preceding and following windows; if two adjacent windows are completely disconnected, the beginning and end of the array will be exposed during reverse addressing, and boundary phase tearing cannot be eliminated.

[0045] Preferably, the time length corresponding to the window length is 0.5 to 0.9 times the first-order low-frequency natural period of the tested entity, more preferably 0.6 to 0.75 times; overlap ratio The value should be between 0.25 and 0.5, more preferably between 0.33 and 0.45. The former ensures that the interior of a single window retains a gradually changing profile under static load, while the latter ensures sufficient window seam connection without excessive memory usage. The relationship between these values ​​is as follows:

[0046]

[0047] Where: Window start time scale : Indicates the window number The corresponding data window is located at the starting point on the absolute time-scale axis, with a positive time value, and its function is to fix the sequence of physical occurrences; window number. Overlapping data window number, a positive integer, used to ensure unique references between adjacent windows; overlap ratio. The proportion of sample points shared between two adjacent windows, ranging from 0.25 to 0.5, serves to provide a boundary junction zone;

[0048] Window length The number of sample points contained in a single window, with a value not less than 32, is used to limit the processing capacity in one operation; sampling interval. The time interval between adjacent samples is a positive time value used to convert time into the number of samples; it is obtained through the sampling system configuration.

[0049] As a supplement: First-order low-frequency natural period Its purpose is to constrain the look-ahead extension length from deviating from the physical boundaries of the measured entity; it is obtained by reading preset values ​​from the structural file, or by averaging the time differences of consecutive unidirectional peaks in the natural attenuation waveform during the power-on calibration phase. Predicted span coefficient. Its purpose is to control the amplification ratio of the extension length relative to the physical period; it is obtained through configuration parameters. Look-ahead extension buffer sequence length. Its purpose is to provide a future end sacrifice region for step two; it is obtained by calculating it using a formula and rounding it down.

[0050] In engineering practice, a dual-zone ring buffer combined with asymmetric ping-pong storage can be used, or a first-in-first-out queue combined with a window index table can be used; both paths include the window start time stamp. Window length The length of the overlapping segment is written to the window header. Therefore, window advancement is unaffected by thread jitter, and there are always common sample points available for intersection at the window seams.

[0051] After completing the overlap capture, the edge gateway does not directly calculate the average across the entire window, but instead extracts the static load baseline only from the central area of ​​the window. Among them, the first and last sample points are close to the truncation boundary, making them prone to incorporating impact sway, bridge switching glitch, or half-wavehead; if these sample points participate in baseline estimation, subsequent analysis from the aliased strain sequence... What is subtracted will no longer be the static load, but the pseudo-static load contaminated by boundary artifacts.

[0052] Preferably, the boundary removal ratio The value should be between 0.08 and 0.18, more preferably between 0.10 and 0.12. The edge gateway should be removed from both ends of the window. Each sample point, extracted only from the median of the sequential statistics in the central region, is expressed as follows:

[0053]

[0054] Where: static load back baseline : Indicates the window number The corresponding data window represents the macroscopic constant load level, with the value being the strain variable, serving as a low-frequency carrier for subsequent reset; aliased strain sequence : Indicates the window number Corresponding sample point number in the data window The corresponding original sample set, with values ​​representing the dependent variable; boundary removal ratio. The percentage of sample points removed from both the beginning and end of the window, ranging from 0.08 to 0.18; window length. The number of sample points contained in a single window, with a value not less than 32, serves to define the upper and lower boundaries of the central sampling area; sample point number. : Indicates the location of the sample point inside a single window, with values ​​ranging from 1 to the window length. The integer, used to index the original sample points and the stripped sample points; the floor operator. : indicates rounding down, which converts the sample point scale boundary into an addressable integer sample point boundary;

[0055] Static load baseline was obtained Afterwards, the edge gateway does not use a single value within the entire window. Instead of performing hard subtraction, transition weights are constructed in the overlapping segments. This causes the previous window to be statically loaded with a baseline. Smooth transition to the current window static load baseline Regenerate transition baseline And obtain pseudo-alternating sequences This approach is to prevent the extraction process itself from creating further steps. It can be expressed as:

[0056]

[0057] Where: Transition weight : Indicates the sample point number within the overlapping segment The corresponding smooth transition coefficient, ranging from 0 to 1, serves to ensure a continuous transition between the baselines of adjacent windows; normalization position : Indicates the sample point number The positional proportion within the overlapping segment, with a value range of 0 to 1, serves to provide a dimensionless abscissa for cubic smooth interpolation. This only indicates the normalized position within the overlapping segment; in step three... It only represents the thermal hysteresis driving force.

[0058] Overlap length The number of sample points shared by the current window and the previous window, with a value not less than 2, is used to limit the effective range of the smooth transition.

[0059]

[0060] Where: transition baseline The baseline value after point-by-point transition within the overlapping segment is taken as the dependent variable and its function is to replace the abruptly changed single-valued baseline in the stripping process; the previous window static load back baseline. : Indicates the previous window number The corresponding data window stores the macroscopic constant load level, with the value being the dependent variable; the current window shows the static load baseline. : Indicates the current window number The corresponding data window stores the macroscopic constant load level, with values ​​representing the dependent variable; pseudo-alternating sequence. : Indicates the window number The corresponding data window is the sequence after baseline stripping, and its value is the dependent variable. Its purpose is to send it to the next step to perform look-ahead extension and bidirectional time-reverse washing.

[0061] In a preferred embodiment, the edge gateway employs a dual-stack selector for median extraction and fixed-point polynomial evaluation for transition weights. The result is a third-order smooth interpolation. If the measured entity is the slewing arm of a port crane gantry, the result directly observable on-site is that the curve that was originally slowly crawling along the high platform is transformed into a thin, elongated band oscillating around the zero line after being stripped. When the previous window is spliced ​​with the current window, there are no longer any visible steps or collapses in the window seams. If the measured entity is changed to a fiber optic grating measuring point, the aliased strain sequence is still used. Static load baseline With pseudo-alternating sequences The same terminology system is used, only replacing the sampling front end with a grating demodulator. Furthermore, it consists of three components working together: central extraction, smooth transition, and baseline stripping, which respectively correspond to the three functions of isolating boundary contamination, reducing transition abrupt changes, and outputting a near-zero mean carrier.

[0062] When static load back baseline With pseudo-alternating sequences After the track separation is completed, the edge gateway carries the static load baseline. Together with the window start time scale Window number The measurement point identifier is written into the temporary storage area, and then the pseudo-alternating sequence is... The data is then sent to the local processing queue for the next step. This temporary storage area participates in subsequent low-frequency resets; preferably, power-down retained memory or ferroelectric memory is used to avoid static load baseline loss after a power failure. Removed from the original timescale.

[0063] To cover outliers caused by welding interference, lightning strikes, or motor start-stop cycles, step one configures an anomaly switching chain. Before central extraction, the edge gateway checks the number of sign flips, the abrupt change in the first-to-last difference, and the bridge saturation flag for samples within the window; if these limits are exceeded, the static load baseline obtained from the current window is not used. Instead, it uses the moving average baseline of the most recent stable window as the alternative stripping baseline and writes the alternative source window number into the registered record. This degradation path does not rewrite the static load back baseline. The main definition is that it only undertakes backup stripping in case of failure. If the acquisition board and the processing board are separate, the acquisition board sets the window start time mark. Window length Overlapping segment length The measurement point identifier and verification field are encapsulated as a JavaScript object representation message JSON header, which is delivered to the computing board via the Transmission Control Protocol (TCP).

[0064] The practicality can be directly observed: comparing unexposed aliased strain sequences on the same screen. The pseudo-alternating sequence after stripping There are still no stepped transitions at the window seams; unfold the static load baseline in the temporary storage area. Does the lifting trajectory match the slow loading steps on site? In this case, in addition to providing the input, output, and execution entity, step one also provides the switching paths for the main solution, alternative solution, and backup solution.

[0065] Thus, step one can stably output two sources of data, namely pseudo-alternating sequences with the original absolute time scale. And the same static load baseline with original absolute time scale The former provides a near-zero mean carrier for the next step of forward extension and bidirectional time-reverse washing, while the latter provides a low-frequency framework for thermal hysteresis stripping and static creep reset.

[0066] Step 2: Utilize the mechanical harmonic constraints of the tested entity itself to generate the pseudo-alternating sequence obtained in Step 1. A physically acceptable future end sacrifice region is added to prevent subsequent forward and inverse filtering from misjudging the truncation boundary as an impact source, and the zero-phase shift alternating sequence within the original window length is retained after the sacrifice region is removed. .

[0067] Step 1: Static Load Baseline From the aliased strain sequence After stripping, pseudo-alternating sequences It already possesses a shape that oscillates around the zero point, but it is still abruptly cut off at the end of the window. As long as the end of the window ends directly, the digital filter with memory length will search for non-existent subsequent samples at the end, and then convert the missing future into boundary distortion. This distortion will roll back to the end along the reverse addressing path, thereby disrupting the subsequent co-location with the static creep reference.

[0068] Therefore, step two does not use the method of extending the sampling waiting window, but instead uses the pseudo-alternating sequence provided in step one. The first-order low-frequency natural period of the measured entity is bound to the edge gateway. The edge gateway first loads the first-order low-frequency natural period from the structure file locally, or backfills the period based on the continuous peak-valley spacing during the power-on calibration phase; then, it constructs a finite-length look-ahead extension buffer sequence only at the end of the window, and then merges the real and virtual samples into the same total sequence. The purpose of this process is to allow the filter to complete the round-trip cleaning within a tamed boundary space, and then truncate the future end after convergence.

[0069] First, the edge gateway uses the pseudo-alternating sequence The initial direction of the look-ahead extension buffer sequence is obtained by determining the evolution direction of the peaks and troughs in the tail section. Next, the extension length is limited by the first-order low-frequency natural period; otherwise, pure mathematical extrapolation would deviate from the mass and damping boundaries of the measured entity. Then, the edge gateway performs forward and reverse filtering with the same coefficients on the total sequence composed of real and virtual samples, causing the phase shifts to cancel each other out during the round trip. Finally, the edge gateway truncates the look-ahead extension buffer sequence along the truncation boundary from step one, obtaining the initial timescale of the window. One-to-one corresponding zero-phase-shift alternating sequence .

[0070] In a representative implementation, the tested entity is a bridge cable. Maintenance personnel lightly tap the cable, and the edge gateway stores the naturally decaying waveform in the structural archive, forming a first-order low-frequency natural period specific to the measurement point. After entering normal operation, the edge gateway retrieves a pseudo-alternating sequence from the queue in step one. The lookahead extension buffer sequence is inserted at the end of the window, followed by forward traversal, backward traversal, and tail removal. The waveform that was originally abruptly broken at the end of the window will be supported by a sacrificial tail that extends forward; after the washing is completed and the tail is removed, the original end of the window will no longer hook back or collapse.

[0071] Before generating the look-ahead extension buffer sequence, the edge gateway does not directly copy the last few samples, nor does it connect the window tails into a horizontal line. Instead, it first copies the samples in the pseudo-alternating sequence. The peak-to-valley distance, the order of adjacent zero crossings, and the direction of the tail slope are examined at the end of the curve to obtain an envelope trend consistent with the current stress state.

[0072] Preferably, the length of the look-ahead extension buffer sequence is controlled by the first-order low-frequency natural period. Edge gateway call window number. The first-order low-frequency natural period of the corresponding measuring point With sampling interval Predicting the span coefficient Locked between 1.25 and 1.5 to obtain the look-ahead extension length. :

[0073]

[0074] Where: look-ahead extension length : Indicates the window number The corresponding data window generates additional virtual sample points at the future end, with values ​​taking positive integers. These points serve to absorb boundary distortion; they also predict the span coefficient. : Represents the expansion ratio of the look-ahead extension buffer sequence relative to the first-order low-frequency natural period, with a value range of 1.25 to 1.5. Its function is to lock the mathematical extrapolation within a structurally acceptable length range; first-order low-frequency natural period : Indicates the window number The first-order low-frequency natural period corresponding to the measurement point is taken as a positive time quantity, and its function is to provide the physical period boundary of the measured entity; sampling interval : Represents the time interval between adjacent samples, with a positive value. Its function is to convert the period length into the number of samples; floor operator. : indicates rounding down, which converts the number of sample points into an addressable integer;

[0075] When the tested entity is a bridge suspension cable, the edge gateway will transmit the first-order low-frequency natural period. Write the data to the measurement point configuration area; when the measured entity is changed to a wind turbine tower, the edge gateway writes the first-order low-frequency natural vibration period obtained from the tap recognition during the startup phase into the same field, and continues to use the first-order low-frequency natural vibration period thereafter. This is the only term. Therefore, the length of the lookahead extension buffer sequence no longer depends on an empirical setting, but rather on the first-order low-frequency natural period. constraint.

[0076] Furthermore, determine the prospective extension length. Subsequently, the edge gateway generates a look-ahead extended buffer sequence at the end of the window. In the optimal primary solution, the edge gateway is based on an autoregressive algorithm using a pseudo-alternating sequence. The tail sample points are used as the initial values ​​for regression, and the resulting look-ahead extended buffer sequence is generated. Recursive backfeeding serves as the input for the next step.

[0077] The resulting virtual wave tail both retains the local phase relationship at the tail end and is also affected by the aforementioned first-order low-frequency natural period. The length limit is defined as follows:

[0078]

[0079]

[0080] Where: look-ahead extension buffer sequence : Indicates the window number Corresponding data window future terminal A virtual sample point, with values ​​set as the dependent variable, serves to provide a filtering sacrifice region; the total sequence is re-fed. : Indicates a pseudo-alternating sequence With look-ahead extended buffer sequence The overall sequence formed by piecing together the first and last parts takes the value of the dependent variable and serves as the input for bidirectional time-reverse washing; the sample point number within the window. : Represents the total recharge sequence The sample point positions in the data range from 1 to... Integer; virtual sample number : Indicates the position number in the lookahead extension buffer sequence, with values ​​ranging from 1 to the lookahead extension length. Integers;

[0081] Regression Index : Indicates the historical sample number referenced by the autoregressive algorithm, with values ​​ranging from 1 to the regression order. Integer values, used to organize local associations at the tail end; window length : Represents the number of single-window sample points determined in step one, with a value not less than 32, serving to delineate the boundary between real and virtual sample points; Regression order : Represents the number of historical samples referenced by the autoregressive algorithm, taking a positive integer value, preferably 4 to 12, more preferably 6 to 10, its function is to balance the preservation of tail shape and computational burden; regression coefficient : Indicates the window number The corresponding data window Each regression weight takes a real number value and is used to describe the local correlation between the tail sample points;

[0082] The length of the tail training segment is denoted as The value ranges from 3p to 6p. The tail training segment starts from a pseudo-alternating sequence. The tail end is truncated forward, and the autocorrelation function is defined as:

[0083]

[0084] in, Indicates the first training segment at the tail end Autocorrelation value; Indicates the length of the training segment; Indicates the window length. The regression coefficients satisfy the Yul-Walker equations:

[0085]

[0086] in, For the regression order; For the first The edge gateway solves the system of equations and writes the regression coefficients into the configuration page of the current window, and then generates a look-ahead extension buffer sequence point by point based on these coefficients. .

[0087] pseudo-alternating sequence : Represents the set of alternating sample points after the static load back baseline has been stripped from the output of step one. The value is the dependent variable, and its function is to provide the true starting point for look-ahead extension.

[0088] In a preferred embodiment, the regression coefficient The solution is obtained by the edge gateway using the Yul-Walker equation of the tail training segment. The solver employs column pivoting or an equivalent linear equation solver. The training segment length is preferably the regression order. Three to six times, to make the regression coefficient Follow the current window tail shape. If the microcontroller issues a memory overflow alarm or junction temperature drops, the fallback solution switches to end-of-line deduction, that is, generating a look-ahead extended buffer sequence along the front with the tail slope, while the interface name remains the look-ahead extended buffer sequence. constant.

[0089] Furthermore, the first-order low-frequency natural period Physical boundary and regression coefficient The local associations are combined and used, with the former constraining the extension length and the latter maintaining the tail texture.

[0090] Obtain the total recharge sequence Then, the edge gateway performs forward filtering and inverse filtering on the total sequence. Forward filtering first traverses the sample points in their natural order, gradually flattening high-frequency noise within the filter's memory length. Inverse filtering then re-feeds the already cleaned sequence to the same set of filter coefficients in reverse time order, canceling out the phase shift introduced during the first traversal on the reverse path. Using the same set of finite impulse response coefficients without changing the other set ensures a closed phase compensation path. Its expression is:

[0091]

[0092] Where: forward washing sequence : Represents the total recharge sequence The sample set after the first sequential filtering, with values ​​representing the dependent variable, serves to provide input for inverse filtering; zero-phase-shift alternating sequence. : Represents the total recharge sequence The output sample set after forward and inverse filtering, with values ​​representing the dependent variable, serves as the alternating carrier ultimately delivered from step two to steps three and four; the sample number within the window... : Indicates a forward wash sequence With zero phase shift alternating sequence The sample point positions in the sequence range from 1 to the total sequence length. The integer is used to locate the corresponding sample points before and after washing;

[0093] Filter coefficients : represents the finite impulse response filter. Each of the following coefficients, taking real numbers, defines the washing bandwidth and transition band shape; the filter coefficients satisfy symmetry and normalization constraints:

[0094]

[0095] in, Indicates the first One filter coefficient; This indicates the filter order. The filter coefficient sequence is pre-stored within the same measurement point configuration area. Forward filtering and inverse filtering use the same set of coefficients, and the coefficient sequence is not switched during operation.

[0096] Filter order : Represents the highest coefficient index of the finite impulse response filter, taking a positive integer value, preferably between 8 and 32, which limits the filter's memory length; Convolution index : Represents the convolution offset between the filter coefficients and the input samples, with values ​​ranging from... Up to filter order The integer, whose function is to complete sequence scrambling; the inverse operator. : Represents a time reversal operation, specifically, it involves reversing the time of a given length. The first input sequence The sample points are mapped to the first... Each sample point serves to construct the inverse filtering path; the total sequence length... : Represents the total recharge sequence The total number of sample points, taking values ​​of Its function is to limit the inverse operator. The index boundary;

[0097] Subsequently, the edge gateway retains only the first part of the original truncation boundary preserved in step one. For each sample point, the entire lookahead extension buffer sequence is truncated. The remaining zero-phase-shift alternating sequence is then... With the window start time scale in step one Window number Maintain a one-to-one correspondence, and also maintain alignment with the static load baseline. The relationship between the same track. The noise spikes after the front window tail will hook back at the tail segment; after washing, the sacrificial wave tail is cut off, and the original window end restores the continuous convergence waveform.

[0098] The forward filtering, inverse filtering, and original boundary pruning are connected as a continuous action. The forward filtering performs the first noise removal, the inverse filtering performs phase compensation, and the original boundary pruning removes the sacrificial region. After the three points are connected, a zero-phase-shift alternating sequence is obtained. Retain the time-scaled anchor point from step one to prepare alternating components for in-situ co-track stitching in step four.

[0099] Extracted by tail envelope tendency and first-order low-frequency natural period Locked-in, the lookahead extended buffer sequence is no longer a mathematical extension detached from physical boundaries, but rather a sacrificial region constrained by structural cycles. This is achieved through regression coefficients. Driven by look-ahead extension, the local peak-valley texture at the window tail is perpetuated, and subsequent filtering no longer directly impacts the hard truncation boundary. This is achieved by analyzing the total recharge sequence. Perform forward and inverse filtering with the same coefficients, then prune the sacrificial region along the original boundary. The zero-phase-shift alternating sequence output in step two is... Maintaining the continuity of the original window length, original window timescale, and alternating information is essential for step three, which involves calling the static load baseline. Thermal hysteresis decoupling and dynamic-static stress alignment in step four provide an alternating input without group delay misalignment.

[0100] Step 3: Remove and temporarily store the static load baseline from Step 1. Reset to the reset static load sequence along the original absolute timescale Then, the thermal migration portion caused by temperature conduction hysteresis is separated from the mechanically stressed creep portion, and a pure static forced creep reference is output for the same track stitching in step four. .

[0101] The static load baseline obtained in step one While macroscopic constant load information is retained, it still contains thermal conduction lag between the metal matrix and the adhesive layer. If the current temperature value is directly subtracted from the table, heat that has not yet entered the strain time will be prematurely deducted, and if the heat remains inside the metal, the deduction will be insufficient, ultimately miswriting mechanical creep as temperature drift pendulum. Therefore, step three incorporates time lag, temperature curvature, and material heat storage into the mechanical-thermal state observation chain simultaneously.

[0102] Step three below is performed by the edge gateway. The temperature channel is provided by a platinum resistance temperature probe or a temperature probe with equivalent functionality. The temperature probe is fixed on the same surface of the measured object as the strain gauge, preferably positioned 5 to 30 mm from the center of the strain gauge grating wire. A thermally conductive adhesive layer is filled between the temperature probe and the measured object, and a windproof cover is placed on the outside. At the bridge box girder site, maintenance personnel press the temperature probe onto the steel plate surface and connect it to the acquisition board; subsequently, the edge gateway receives the static load baseline at an absolute time scale. The temperature sampling values ​​are displayed on the screen, and two curves can be seen unfolding along the same axis.

[0103] Among them, the edge gateway first determines the window start time stamp. Window length Combined with the handover record saved in step one, the static load baseline is set. Backfill is a reset static load sequence corresponding to each sample point. Then, the ambient temperature sequence Resampling to and reset static sequence A perfectly consistent time grid, thereby generating thermal hysteresis driving forces. Next, the edge gateway sends the reset static sequence. As an observation input, the thermal hysteresis driving force As the state excitation, a Kalman force-thermal multidimensional state observer is run; finally, a pure static forced creep benchmark is projected from the state vector. The residual amount of hot-moving data is left in the local cache and is not included in the same-track summation in step four.

[0104] In the example, the tested entity is a wind turbine tower flange section, and the zero-phase shift alternating sequence is used. It exists in the edge gateway's memory. The edge gateway does not store the zero-phase-shift alternating sequence. Instead of directly applying the static load, the high-frequency vibration intensity is converted into the structure's natural energy level for use in the penalty gating in the alternative scheme. This arrangement ensures that step three always revolves around the static load reset chain, preventing the separated alternating components from being fed back into the low-frequency channel.

[0105] The edge gateway first sets the static load baseline. Reset to reset static load sequence The specific procedure is to follow the window number from step one. Window start time marker And overlapping handover records, put the static load baseline of each window. Write back the sample interval; if the sample is located in the overlapping area of ​​two adjacent windows, then continuously backfill along the smooth transition direction adopted in step one, thereby avoiding the step generated by the reset action itself. After the reset is completed, the edge gateway projects the temperature channel sample values ​​onto the same sample time scale using either final value hold or linear interpolation to obtain the ambient temperature sequence. .

[0106] Subsequently, the edge gateway does not directly use the ambient temperature sequence. Instead of the current value, it constructs the thermal hysteresis driving force. This quantity simultaneously includes the temperature hysteresis, the second-order temperature change, and the specific heat capacity storage term. It is used to describe three different physical states: heat is still on its way, heat is accelerating into the material, and heat has already been stored within the material. Its expression is:

[0107]

[0108] Where: thermal hysteresis driving force : Indicates the window number Sample point number The comprehensive thermal drive caused by temperature evolution, taking real values, serves as the excitation input for the mechanothermal state observation; its purpose is to serve as the excitation input for the state observer; it is obtained from the ambient temperature sequence. The calibrated parameters are calculated using the formula.

[0109] Ambient temperature sequence : Indicates the window number Sample point number The temperature value after time-scale alignment is a temperature quantity and serves as the original source of thermal hysteresis drive; hysteresis interval. : Represents the number of discrete delay samples for heat propagation from the surface of the measured object to the strain gauge wires. The value is a positive integer, preferably 1 to 12, and its function is to characterize the heat conduction hysteresis; sampling interval. : Represents the time interval between adjacent sample points, with a positive time value. Its function is to convert the second-order temperature change into a curvature term per unit time.

[0110] Thermal inertia window : Represents the number of sample points for the cumulative summation of heat storage, taking a positive integer from 4 to 32. This term represents the degree of heat retention within the material; specific heat capacity. Window number The specific heat capacity term of the corresponding physical material is taken as a positive real number; hysteresis weight. : Represents the proportional coefficient of the thermal hysteresis term, taken as a real number, and the weight of the delayed temperature difference term, representing the influence of temperature hysteresis on static load drift; obtained by least squares fitting of the thermal calibration segment.

[0111] Curvature weight : Proportional coefficient of the second-order temperature variation term, taken as a real number, used to measure the impact of temperature abrupt curvature on the thermal path; weight of the second-order temperature variation term; used to measure the impact of temperature curvature on hysteresis; used to obtain the least squares fit of the thermal calibration segment; heat storage weight. : Specific heat capacity storage term proportionality coefficient, taken as a real number, used to measure the drag of material heat storage on static load travel; weight of heat storage term; used to measure the cumulative effect of thermal inertia; summation index. : Indicates the sample point number within the thermal inertia window, with an integer value, and its function is to complete heat storage accumulation;

[0112] First, reset the static load sequence. and ambient temperature sequence Unify to the same time scale, and then generate thermal hysteresis driving quantities. Subsequent observers no longer receive a single temperature value, but rather a complex excitation carrying propagation hysteresis, temperature curvature, and thermal memory.

[0113] Thermal hysteresis driving force After construction, the edge gateway will reset the static load sequence. The data is fed into a Kalman force-thermal multidimensional state observer. The state vector of this observer not only contains a single low-frequency quantity, but also simultaneously contains the pure static forced creep component, the thermal wandering residual, and the thermal inertial memory.

[0114] Among them, the pure static forced creep component is responsible for the slow migration under real mechanical load, the thermal migration residual is responsible for the apparent drift affected by temperature, and the thermal inertia memory is responsible for retaining the historical heat that has not yet been released.

[0115] In the optimal primary scheme, the edge gateway utilizes the reset static load sequence after backfilling in step one. As an observable quantity, the thermal hysteresis driving force is utilized. As input, the state is advanced according to the following relationship:

[0116]

[0117] Where: Predicted state vector : Indicates the window number Sample point number The predicted force-thermal state values ​​before observational correction are represented as a column vector, serving to connect the state evolution of the previous point in time; force-thermal state vector. : Indicates the window number Sample point number The force-thermal state after observation correction is a column vector, which simultaneously contains the pure static forced creep component, thermal migration residual, and thermal inertia memory; it is a force-thermal state vector; its purpose is to bear the three types of states: mechanical creep, thermal migration, and thermal inertia; it is obtained by recursion.

[0118] State transition matrix : Indicates the window number The corresponding state evolution matrix is ​​a real matrix, used to describe the migration of mechanical creep and thermal inertia; the state transition matrix is ​​used to describe the internal evolution and coupling of the three types of states; it is obtained by writing the calibration parameters of the same measurement point.

[0119] Drive matrix : Indicates the window number The corresponding thermal drive coupling matrix is ​​a real matrix, and its function is to convert the thermal hysteresis drive quantity into a real matrix. Injected state vector; correction gain matrix : Indicates the window number The corresponding observation correction matrix, which is a real matrix, is used to correct the predicted state vector based on the observation residuals. ; Reset static load sequence : Indicates that the baseline is supported by static load. The sample-level static load sequence obtained by backfilling along the original timescale is used as the dependent variable and serves as the observation input; the observation matrix... : Indicates the window number The corresponding observation mapping matrix, which is a real matrix, is used to map the predicted state vector. Projection is a reversible static load sequence Comparison of observations;

[0120] In a preferred implementation, the state transition matrix A three-dimensional state is adopted, with the first dimension corresponding to the pure static forced creep component, the second dimension corresponding to the thermal migration residual, and the third dimension corresponding to the thermal inertial memory. Matrix solving employs a square root Kalman path or an equivalent function matrix decomposition path. If a parallel alternative is implemented in the field, the edge gateway is replaced with a lightweight long short-term memory network; its input tensor receives the reset static load sequence along the same terminology system. Ambient temperature sequence Thermal hysteresis driving force and the zero-phase-shift alternating sequence The local energy is converted into the structural natural energy level and added to the forget gate weights, suppressing the network from mistaking the mechanical natural wake as thermal drift.

[0121] Furthermore, the observer does not passively subtract the current temperature value, but converges back and forth between state transitions and observation corrections. The pure static forced creep component, thermal wandering residual, and thermal inertial memory are carried by different state channels, and the contamination of static load by thermal hysteresis is compressed into a separate channel.

[0122] After completing the force-thermal state observation, the edge gateway starts from the force-thermal state vector. Projecting the pure static forced creep benchmark In the optimal master scheme, the projection matrix only reads the state components bearing mechanical creep, and does not read the thermal wandering residual and thermal inertia memory, so that the output only retains the gradual change of force and does not carry thermal drift. Its expression is:

[0123]

[0124] Where: Pure static forced creep reference : Indicates the window number Sample point number The static force reference after thermal hysteresis decoupling is set to the strain variable, and its function is to serve as the static component input for the same-track stitching in step four; projection matrix : Indicates the window number The corresponding state selection matrix, which is a real matrix, is used to select the state from the force-thermal state vector. Extract the mechanical creep component; mechanical and thermal state vector. : Represents the multidimensional mechanical and thermal state after observation correction. It takes the value of a column vector and serves to provide a separation carrier for the static and thermal components.

[0125] In one specific implementation, the tested entity is a large rolling mill column. The operator first attaches the strain gauges and temperature probes, and then enters the specific heat capacity of the column material into the edge gateway. And the probe installation location. The reset static load sequence displayed on the screen during equipment heating. The overall rise will be slow and gradual; after step three is completed, the output pure static forced creep benchmark will be obtained. The slow migration along the compressive direction is maintained, while the bulging drift that occurs synchronously with the temperature rise is left in the thermal walk buffer. If a memory overflow alarm or junction temperature drop occurs, the degradation backup scheme no longer runs matrix observations, but instead calls a one-dimensional sliding time window lookup table containing a fixed empirical time delay constant; the lookup table still uses the ambient temperature sequence. and reset static load sequence As input, with a pure static forced creep benchmark. This provides the output, thus maintaining an input / output interface consistent with the main scheme.

[0126] Furthermore, it includes the primary solution, alternative solutions, and a safety net solution. The primary solution depends on the projection matrix. From the mechanical and thermal state vector The mechanical creep component is extracted; an alternative solution incorporates the structure's natural energy level into network gating; a fallback solution retains the input / output interfaces. Specifically, the pure static forced creep benchmark before and after the abrupt temperature change can be directly observed. Whether it continues to extend along the direction of the load.

[0127] By using the static load baseline Backfill into reset static load sequence and the ambient temperature sequence Unified to the same time scale, using thermal hysteresis driving force By incorporating temperature hysteresis, temperature curvature, and specific heat capacity into the same driving chain, thermal expansion migration has an observable entry point.

[0128] Then, using a Kalman force-thermal multidimensional state observer or an alternative path with the same interface, step three extracts the pure static forced creep component from the thermal migration residual, outputting the pure static forced creep baseline. With the zero-phase-shift alternating sequence of step two Maintaining a clear division of labor based on time scale, different frequency bands, and different physical meanings provides a temperature-drift-free static component for the in-situ co-track stitching in step four.

[0129] Step 4: Transform the zero-phase-shift sequence Compared with pure static forced creep benchmark Stitching the data point by point on the same absolute time scale into a full-band synchronous stress sequence The sequence was then fed into the edge-side and cloud-side rainflow damage closed loop chain.

[0130] Step two has yielded the zero-phase-shift alternating sequence. Step three has yielded a pure static forced creep baseline. However, the two come from different computation chains. If the output is only added in the order of arrival of the cache, data with different window numbers and different point numbers will be forcibly concatenated to the same moment, thereby creating pseudo half-cycles and pseudo closed loops in the peak-valley chain, and the rainflow damage matrix will be distorted accordingly.

[0131] Step four below is performed by the edge gateway; the edge gateway includes a real-time clock module, double-buffered static random access memory, and an Ethernet port or equivalent wireless transmission port. Step two involves writing the zero-phase-shift alternating sequence into the output queue. Step 3: Write the pure static forced creep baseline to the output queue. Step four only extracts the sequence number of the same window. Same point number The sample points.

[0132] In a representative implementation, the tested entity is a bridge main cable suspender. On the local screen, two curves to be stitched are first judged by the time matching indicator light and then enter the same page cache; sample point pairs that fail the judgment enter the mismatch recording area. Thus, time consistency is pre-fixed.

[0133] Step four unfolds along the single chain of same-track determination - linear stitching - closed-loop output. The edge gateway first determines the window start timescale. Window length With sampling interval Zero phase shift alternating sequence and pure static forced creep benchmark The sample-level time index is reconstructed, and the synchronization relationship is verified within the nanosecond-level synchronization dead zone. Then, the verified sample pairs are linearly stitched point by point, and stress conversion is completed according to the strain-stress conversion coefficient of the measurement points to obtain the full-band synchronization stress sequence. Finally, the edge gateway will synchronize the stress sequence across the entire frequency band. One path is written to the local rainflow closed-loop chain, and the other path is encapsulated into a message and sent to the cloud group control center server.

[0134] Time verification only performs index checks and dead zone determination; linear stitching only performs same-track addition and stress conversion; closed-loop output only performs peak and valley extraction, matrix accumulation, and alarm linkage.

[0135] Before performing sample summation, the edge gateway first performs a zero-phase-shift alternating sequence. Compared with pure static forced creep benchmark Attachment time. The attachment method is not resampling, but rather based on the window start timescale retained in step one. With sampling interval Recover the theoretical occurrence time of each sample point, and then add the residual path bias formed when each output queue is written to the buffer.

[0136] For example, zero-phase-shift alternating sequence Cache pages and pure static forced creep benchmark The cached pages are simultaneously loaded into the same dual-port static random access memory. The edge gateway first reads the header fields of both pages, including the window number. Window start time marker Window length Sampling interval The offset from the first sample point within the page; the point-by-point verification within the page is only initiated when it is consistent with the page header.

[0137] The verification relationship is as follows:

[0138]

[0139] Where: zero-phase-shift alternating timestamp : Sample time of alternating components, the value is a time quantity, used for same-track verification; Static reference timestamp : Sample time of static components, the value is a time quantity, used for track calibration; Window start time scale The absolute starting point of the current data window, with a time value, used to unify the starting point of the two data streams;

[0140] Sample number : The location of the sample point inside the single window, with values ​​ranging from 1 to the window length. Integers are used to establish a one-to-one correspondence; sampling interval : The time interval between adjacent samples, a positive time value, used to map the sample time; Alternating path offset The residual path offset generated when the alternating channel is written to the buffer is a time-based value used to compensate for the arrival difference of the alternating channel; static path offset. The residual path offset generated when writing to the cache from a static channel is a time-based value used to compensate for the arrival difference of the static channel.

[0141] Subsequently, the edge gateway performs a synchronization dead zone determination for each sample pair:

[0142]

[0143] Where: synchronization difference The absolute difference between the times of the two sample points, taken as a non-negative time value, is used to determine whether a sample point is allowed to enter the same track for stitching; zero phase shift alternating timestamp. : Alternating component sampling time, the value is a time quantity, used to calculate synchronization difference. Static base timestamp : Static component sampling time, the value is a time quantity, used to calculate the synchronization difference. Synchronization tolerance : Upper limit of synchronization dead zone, which is a positive time value, preferably in the nanosecond to microsecond range, to shield buffer arrival jitter;

[0144] If the synchronization difference Exceeding synchronization tolerance The edge gateway does not borrow samples across windows, nor does it pull samples from the end of the previous page to fill gaps. Instead, it writes the current sample pair into the mismatch record area and continues to search for the next set of sample pairs that pass the judgment along the same window. Thus, the mismatch is limited to a local area.

[0145] Once the sample pair passes the synchronization dead zone determination, the edge gateway performs zero-phase-shift alternating sequence processing. Compared with pure static forced creep benchmark Perform point-by-point suturing. This suturing is neither cross-window splicing nor segmental replacement, but rather suturing within the same window sequence number. Same point number Linear summation is then performed to bring the alternating component separated in step two and the static component separated in step three back to the same physical strain trajectory; subsequently, stress conversion is completed based on the strain-stress conversion coefficient corresponding to the measuring point.

[0146] In a preferred embodiment, the tested entity is a steel boom, with strain gauges attached to its surface. Material elastic modulus and calibration section conversion information are pre-written into the edge gateway, forming a strain-stress conversion coefficient specific to each measuring point. The high-frequency oscillation caused by wind load is a zero-phase-shift alternating sequence. The concepts of load bearing, constant load, and slow creep are derived from the pure static forced creep benchmark. The two components, after being linearly stitched together, simultaneously retain both rapid fluctuations and an overall upward tilt in the waveform. Their relationship is as follows:

[0147]

[0148] Where: same-track synthetic sequence The total strain sequence after stitching at the same time point, with values ​​representing strain, is used for stress conversion; zero-phase-shift alternating sequence. The alternating component output in step two is the strain, used to provide a rapid stress change reference; pure static forced creep reference. The static component output in step three, whose value is the strain, is used to provide a low-frequency stress reference; the full-frequency synchronous stress sequence. : The output sequence after stress conversion, with values ​​representing stress, is used for external interfaces and rainflow closed loops;

[0149] Strain-stress conversion factor The strain-stress conversion parameters corresponding to the measuring points are positive real numbers, used to synthesize the same-track sequence. Mapped to full-band synchronous stress sequence In-situ synthetic sequences The total strain sequence after stitching, with values ​​representing the strain, is used in stress conversion.

[0150] For example, if the tested entity is changed to an aluminum alloy track beam, then the strain-stress conversion factor... Replace with the material and configuration calibration values ​​corresponding to that measurement point; the same terminology system still maintains the synchronous stress sequence across the entire frequency band. This is the only output designation.

[0151] Furthermore, the same-track synthetic sequence Synchronous stress sequence across the entire frequency band without introducing additional phase shift It corresponds point-by-point to the original physical time.

[0152] Obtain the full-band synchronous stress sequence Subsequently, the edge gateway immediately performs peak-valley chain extraction and rainflow pairing locally, and simultaneously encapsulates outgoing messages. The message header typically includes the measurement point identifier and window sequence number. Window start time marker Sampling interval Window length Alarm status word, cyclic redundancy check field; message body contains full-band synchronization stress sequence. The sample array, peak-valley chain index, and incremental rainflow results for this window are obtained. The edge gateway serializes the message into a JavaScript object representation message (JSON), which is then sent to the cloud-based group control center server via Message Queuing Telemetry Transport Protocol (MQTT). After receiving the message, the server processes it according to the window sequence number. Remove duplicates and refresh the digital twin fatigue page for the measurement points.

[0153] In the local rainflow closed loop, the edge gateway no longer waits for cloud feedback before counting; it directly uses the full-band synchronous stress sequence. The current window's rainflow increment matrix is ​​directly generated and superimposed with the cumulative damage matrix, as follows:

[0154]

[0155] Where: cumulative damage matrix Amplitude box number With mean bin number The cumulative damage at a given point, taking non-negative real numbers, is used to record the fatigue accumulation state; the cumulative damage matrix of the previous window. The cumulative damage amount corresponding to the previous window, with a non-negative real number value, is used to provide the cumulative basis; the rainflow increment matrix for this window. The number of newly added rainflow circulation meters in this window, with values ​​being non-negative real numbers, is used to record the newly added circulations in this window; Damage weight matrix. Single-cycle damage weight, a non-negative real number, used to convert the number of cycles into damage increment; amplitude bin number. Box numbering based on stress amplitude, with values ​​being positive integers, used to locate different amplitude levels; average value box number. : Box location numbering based on average stress value, with a positive integer value, used to locate different average value levels;

[0156] In a field network outage implementation, if the cloud-based group control center server does not return an acknowledgment, the edge gateway stores the JavaScript object representation message JSON in a local circular cache, does not delete the current window's rainflow increment matrix, and continues to refresh the cumulative damage matrix. When the cumulative damage matrix corresponding to a certain bin When a preset threshold is triggered, the edge gateway sends a hardware emergency stop level pulse to the programmable logic controller (PLC) via Modbus bus messages or hardware input / output interrupt lines. On-site personnel can directly observe this: the peak-valley chain on the local screen continuously scrolls, the communication status light turns to offline color, and the control cabinet alarm light illuminates with the emergency stop level.

[0157] Furthermore, edge closed-loop, cloud backhaul, and network outage backup are combined into an output chain. Message backhaul continuously updates the cloud-side digital twin, and the local cumulative damage matrix... Fatigue assessment can still be performed even after continuous network outages, and continuous hardware emergency stop pulses can protect against network disconnection.

[0158] By determining the absolute timescale alignment, step four transforms the zero-phase-shift alternating sequence. and pure static forced creep base Quasi-constraints are applied simultaneously, followed by linear stitching to prevent spurious peaks and valleys caused by cross-window sampling and cross-cohort mismatches. This results in the synthesis of sequences along the same track. Sequential application of strain-stress conversion factors, full-band synchronous stress sequence The morphology of both alternating and static loads is preserved, and group delay is not reintroduced.

[0159] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0160] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0161] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0162] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0163] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for simultaneous measurement of dynamic and static stress based on strain gauges, characterized in that: Performed by the edge gateway, including: Obtain the aliased strain sequence with absolute timestamps and truncate it according to the overlapping time window; extract the extremely low frequency static load negative baseline for each time window and temporarily store it, and subtract the static load negative baseline from each time window to form a pseudo-alternating sequence. Based on the tail envelope trend of the pseudo-alternating sequence and combined with the dominant frequency of the tested entity structure, a look-ahead extended sequence is generated. The look-ahead extended sequence is spliced ​​to the tail of the pseudo-alternating sequence to form a total sequence. The total sequence consists only of the pseudo-alternating sequence and the look-ahead extended sequence. The total sequence is subjected to bidirectional time-inverse filtering and the look-ahead extended sequence is removed to obtain a zero-phase-shift alternating sequence. The static load back baseline is reset, and the temperature sequence is input into the hysteresis constraint state space model to separate thermal drift and mechanical creep, thus obtaining the static creep baseline. The zero-phase-shift alternating sequence is superimposed on the static creep reference along the same track according to the absolute timestamp, and the synchronous stress sequence is output.

2. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 1, characterized in that: The edge gateway extracts the static load baseline in the central area after removing the first and last boundary samples in each time window. The static load negative baseline of the previous time window is transitioned to the static load negative baseline of the current time window according to the sample point number for the overlapping area shared by adjacent time windows, forming a transition baseline; the mixed strain sequence is stripped point by point according to the transition baseline to form a pseudo-alternating sequence.

3. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 2, characterized in that: The edge gateway extracts the training segment from the tail of the pseudo-alternating sequence and generates a look-ahead extension sequence based on the tail envelope trend. The pseudo-alternating sequence in the total sequence is formed by sequentially splicing the pseudo-alternating sequence of the overlapping area of ​​the previous time window and the pseudo-alternating sequence of the current time window. Forward filtering and inverse filtering after time reversal are performed on the total sequence according to the same set of digital filtering coefficients. Only the sample points within the boundary of the current time window are retained as the zero-phase-shift alternating sequence.

4. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 1, characterized in that: The edge gateway backfills the static load backing baseline into a reset static load sequence according to the absolute timestamp, and aligns the temperature sequence with the reset static load sequence to the same grid; inputs the reset static load sequence and temperature sequence into a hysteresis constraint state space model containing mechanical creep state, thermal wandering residual state and thermal inertia memory state; extracts the static creep baseline from the output state of the hysteresis constraint state space model.

5. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 4, characterized in that: The edge gateway superimposes the zero-phase-shift alternating sequence and the static creep reference on the same track according to the absolute timestamp, including: writing the time window number, time window start timestamp, time window length and sampling interval into the page header of the zero-phase-shift alternating sequence and the page header of the static creep reference, respectively; reconstructing the time index of the two sample points based on the page header fields; and performing the same track superposition only on sample point pairs whose time difference is not greater than the synchronization tolerance.

6. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 5, characterized in that: When the edge gateway outputs the synchronous stress sequence, it encapsulates the synchronous stress sequence into a data frame. The frame header of the data frame includes the measurement point identifier, time window number, time window start timestamp, sampling interval, and alarm status word. The frame body includes the sample point array of the synchronous stress sequence and the corresponding peak and valley indexes, and is output continuously in page order.

7. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 5, characterized in that: The edge gateway establishes a mismatch record area for sample pairs that fail the synchronization tolerance judgment; the mismatch record area records the time window number, sample number, zero phase shift alternating sequence sample time index, static creep reference sample time index, and sample time difference. The edge gateway skips the sample point pairs corresponding to the mismatch record area within the current time window and does not write them into the synchronous stress sequence.

8. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 1, characterized in that: The length of the overlapping time window is 0.5 to 0.9 times the first-order low-frequency natural period of the tested entity, and the overlap ratio of the overlapping time window is 0.25 to 0.

5. After the edge gateway excludes the first and last boundary samples in the central area of ​​each time window according to the predetermined boundary elimination ratio, it then extracts the extremely low frequency static load back baseline.

9. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 8, characterized in that: The edge gateway determines the length of the look-ahead extension sequence according to the first-order low-frequency natural period of the tested entity and the sampling interval. The time span corresponding to the look-ahead extension sequence is 1.25 to 1.5 times the first-order low-frequency natural period. The edge gateway extracts a training segment with a length of 3 to 6 times the regression order from the end of the pseudo-alternating sequence to generate the look-ahead extension sequence.

10. The method for simultaneous measurement of dynamic and static stress based on strain gauges according to claim 1, characterized in that: The edge gateway includes a real-time clock module, a double-buffered static random access memory, and a direct memory access scheduling unit; The direct memory access scheduling unit writes adjacent time windows to two memory areas during half-transfer interrupts and full-write interrupts, respectively. The edge gateway alternately performs static back baseline extraction, look-ahead extension sequence generation, and co-track overlay between two memory regions.

Images